Human-first AI transformation: beyond getting it right

Let's talk about the first Asian bank to be the subject of a Harvard Business Review case study: DBS. (It was also named the World's Best AI Bank 2025 by Global Finance.)

This case study hits pretty close to home for the Collar founders. We're Singaporean, and many of us from this country had DBS as our first ever bank account.

DBS answers the perpetual question every non-technical person asks: what happens to the people inside the organisation when AI is implemented at scale? DBS is Southeast Asia's largest bank by assets and employs over 40,000 people across 19 markets.

Slow, steady AI investment

DBS began secretly building their data and AI infrastructure in 2014 — years before generative AI entered the mainstream conversation. This early start meant that by the time the rest of the industry was scrambling to deploy LLMs, DBS had already built the data foundations, governance frameworks, and internal capability to move with confidence.

The results by 2025 speak to the compounding effect of that patience:

• 1,500+ AI models deployed across 370+ use cases
• SGD 1 billion in economic value generated from AI initiatives in 2025 — a target the bank publicly committed to and hit
• Named World's Best AI Bank by Global Finance, the first Asian institution to receive the recognition
• 9 Operating Model Transformation (OMT) initiatives completed in 2025, surpassing the original target of 6

The human-first philosophy

The defining characteristic of DBS's approach is not what it deployed — it is how it thought about the relationship between AI and its people. Where many organisations treat workforce impact as a downstream consideration, DBS built it into the strategy from the start.

The bank's position is explicit: AI is an augmentation tool, not a replacement programme. Every AI deployment is designed around the question of how it changes what a human does — not whether the human is still needed. This is not a PR position. It is an architectural constraint that shapes how use cases are selected, how models are deployed, and how success is measured.

"We are not deploying AI to reduce headcount. We are deploying AI to change what our people spend their time on." — Reflective of DBS's public posture on workforce and AI, 2024–2025

Rebuilding the workforce, not just the tech stack

DBS's commitment to its people is not rhetorical. It is operational. The bank identified 13,000 employees for comprehensive upskilling in future-ready skills including AI and data analytics, with over 10,000 already on active learning roadmaps by 2025. An additional 12,000 staff were earmarked for specialised reskilling in the same year.

Rather than treating this as a training programme, DBS treated it as an organisational redesign. Teams were restructured. Roles were redefined. The bank committed to creating approximately 1,000 new AI-related positions to absorb the capacity freed up by automation.

The bank also built iCoach — a generative-AI-powered career coaching platform developed with executive coaching pioneer Marshall Goldsmith. iCoach gives every DBS employee access to personalised, on-demand career guidance 24 hours a day, drawing on the bank's own internal role definitions, mobility pathways, and learning resources. The message it sends to staff is direct: we are investing in your future, not just the bank's efficiency.

The PURE framework: responsible AI as competitive advantage

While most organisations treat AI governance as a compliance obligation, DBS treated it as a strategic asset. The bank developed the PURE framework — four principles that govern how AI is used internally and in customer-facing applications:

PURE is not a set of guidelines posted on an intranet. It is an operational standard embedded into the AI development lifecycle. Every new AI use case is evaluated against it before deployment. This has two effects: it builds employee confidence in the tools they use (because they know the tools have been tested against clear ethical standards), and it builds customer trust (because AI behaviour is predictable and explainable).

In an environment where regulators across Asia are increasing scrutiny of AI in financial services, DBS's governance posture has become a competitive differentiator, not just a risk mitigation strategy.

Operating Model Transformation: redesigning work, not just automating it

DBS's most distinctive organisational contribution is the concept of Operating Model Transformation (OMT). Rather than simply deploying AI tools on top of existing processes, OMT involves fundamentally rethinking how a team's work is structured around human-AI collaboration.

Each OMT initiative involves three elements: rebuilding the workflow itself (not just adding AI to it), upskilling the people in that workflow, and restructuring the team around the new division of labour between humans and AI. In 2025, DBS completed nine such transformations across its major business lines. Specific outcomes include:

• The CSO Assistant GenAI tool reduced call handling times by up to 20%, freeing customer service teams for complex, high-value interactions
• A GenAI assistant now handles 250,000 monthly customer queries — automating 80% of routine tasks so human agents can focus on the 20% that requires judgment
• CodeBuddy, an internal GenAI tool for software developers, cut test case generation and documentation timelines from months to weeks
• An AI-powered risk scoring model now vets 100% of technology change requests, producing an 81% reduction in system incidents caused by those changes
• GenAI tools have delivered an estimated 5–10% time savings across day-to-day work for 40,000+ employees

The human-first thesis

The DBS story yields a different set of lessons from the JP Morgan story — not contradictory, but complementary. Where JP Morgan demonstrates what happens when you invest at scale and deploy fast, DBS demonstrates what happens when you invest in people as deliberately as you invest in technology.

Lesson 1: the workforce transition is the transformation

Most organisations treat workforce transition as a side effect of AI deployment. DBS treats it as the primary objective. The question guiding every use case is not "can AI do this?" but "what does this free our people to do instead?" This reframe changes what you build, how you measure success, and how you communicate the programme internally. It also changes the adoption curve: when employees understand that AI is changing what they do rather than removing them from the equation, resistance falls and engagement rises.

Lesson 2: governance is a trust-building instrument

The PURE framework is DBS's most exportable innovation. Not because the four principles are novel, but because DBS operationalised them. They are not a values statement. They are a development standard. Organisations that bolt governance onto AI programmes after deployment consistently face the same problem: employees distrust tools they do not understand, and customers distrust AI decisions they cannot explain.

Lesson 3: patience compounds

DBS's 2025 results did not emerge from a single large programme. They emerged from eleven years of consistent investment in data infrastructure, AI capability, and human skills. This is not an argument for moving slowly. It is an argument for investing in the right foundations from the start, rather than chasing short-term deployment metrics that do not compound.

Lesson 4: Operating Model Transformation is the real prize

The highest-value AI use cases are not in individual tool deployments. They are in the redesign of how entire teams work. A single GenAI tool that saves 10% of one person's time is a productivity improvement. A team restructured around human-AI collaboration, with roles redefined and workflows rebuilt, is a step-change. DBS completed nine such redesigns in 2025. Most organisations have not completed one.

Lesson 5: trust is the scalability constraint

Speed of AI deployment is not the primary constraint on value creation. Trust is. Employees who do not trust AI tools use them superficially or revert to manual processes after pilots end. DBS built its transformation around trust — with employees through transparent upskilling investment and iCoach, with customers through the PURE framework, with regulators through explainable governance. This is why adoption ran deep, not just wide.

Two models, one destination

JP Morgan and DBS both achieved transformational AI outcomes. They took different paths. Understanding both is more useful than treating either as the single right answer.

Neither model is superior. JP Morgan's approach works in large, fast-moving financial institutions with the capital to deploy at scale and the governance infrastructure to manage risk. DBS's approach works in organisations that operate in multiple regulated markets, need deep employee trust, and are willing to invest in the long arc of transformation.

For most mid-sized organisations, DBS's path is the more replicable one: start with the data foundation, build governance into the process from day one, invest in your people as deliberately as you invest in your models, and treat operating model redesign as the real deliverable.

What the DBS model means in practice

Diagnose before deploying. DBS's OMT framework starts with a diagnosis of how work is currently structured and where AI creates genuine leverage. This is not a technology audit — it is a workflow audit.

Invest in the people equation explicitly. DBS did not assume employees would figure out how to work with AI on their own. It invested in structured upskilling, identified individuals for reskilling, built iCoach, and committed to creating new roles. The investment in people was proportional to the investment in technology — and it is what made the technology investments stick.

Build governance before you need it. PURE was built before DBS hit the regulatory scrutiny that now accompanies AI deployment in financial services. That sequencing mattered. Organisations that build governance frameworks in response to incidents are always playing catch-up.

Measure the right things. DBS measured economic value generated from AI — a hard number that required attributing business outcomes to specific deployments. The SGD 1 billion target, publicly committed to and hit, was the anchor that kept the programme aligned to business reality rather than AI theatre.

Summary

JP Morgan shows what AI transformation looks like at maximum velocity and scale. DBS shows what it looks like when you get the foundations right and invest in your people with the same rigour you invest in your technology.

The DBS story is not about being slower or more cautious than JP Morgan. It is about recognising that the hardest part of AI transformation is not deploying models — it is changing how 40,000 people work, and doing it in a way that earns their trust rather than eroding it.

That is what the PURE framework is for. That is what iCoach is for. That is what Operating Model Transformation is for. And that is why DBS is the World's Best AI Bank — not because it deployed the most AI, but because it built the kind of organisation that uses AI well.

The organisations that will win the next decade of AI are not the ones that deploy the fastest. They are the ones that build the deepest — in their data, their governance, their people, and their operating models. DBS is the proof of concept.

Sources

• DBS Named World's Best AI Bank 2025 (DBS Newsroom): dbs.com/newsroom
• Harvard Business School: DBS AI Journey Case Study: hbs.edu
• DBS AI-Powered Digital Transformation (DBS): dbs.com/artificial-intelligence
• DBS Rewires Operating Models for AI Reasoning Era (Computer Weekly): computerweekly.com
• DBS Billion-Dollar AI Dream Realized (Forrester): forrester.com
• DBS AI Strategy Analysis (Klover.ai): klover.ai
• DBS iCoach GenAI Workforce Tool (DBS Newsroom): dbs.com/newsroom/icoach
• DBS Human-AI Synergy Approach (Tearsheet): tearsheet.co